Application of artificial neural networks to the analysis of two‐dimensional fluorescence spectra in recombinant E coli fermentation processes

Abstract

A two-dimensional (2D) spectrofluorometer was used to monitor various fermentation processes with recombinant E coli for the production of 5-aminolevulinic acid (ALA). The whole fluorescence spectral data obtained during a process were analyzed using artificial neural networks, ie self-organizing map (SOM) and feedforward backpropagation neural network (BPNN). The SOM-based classification of the whole spectral data has made it possible to qualitatively associate some process parameters with the normalized weights and variances, and to select some useful combinations of excitation and emission wavelengths. Based on the classified fluorescence spectra a supervised BPNN algorithm was used to predict some of the process parameters. It was also shown that the BPNN models could elucidate some sections of the process's performance, eg forecasting the process's performance.

In situ 2D fluorometry and chemometric monitoring of mammalian cell cultures

“…This very important validation step should be ''tough'' by exposing the final model to a wide range of completely new measurements. A popular strategy is splitting randomly the measured dataset into a calibration dataset (with $70% of the measurements) and a validation data set (with $30% of the measurements) (Lee et al, 2005;Rhee and Kang, 2007). With this strategy, data interpolation should be avoided; otherwise the probability of similar measurements being inserted in the validation and calibration data sets is very high.…”

“…Partial least squares (PLS) regression appears to be the most popular chemometric method for calibrating 2D fluorescence maps with off-line bioprocess variables such as biomass, substrates and products of interest (Boehl et al, 2003;Lantz et al, 2006;Surribas et al, 2006b;Haack et al, 2007;Rhee and Kang, 2007). Due to the inherently nonlinear nature of biological processes, some authors have used artificial neural networks (ANN) (Wolf et al, 2001;Lee et al, 2005) and nonlinear PLS as calibration methods (Lee et al, 2006).…”

On-line estimation of biomass, glucose and ethanol in Saccharomyces cerevisiae cultivations using in-situ multi-wavelength fluorescence and software sensors

“…Successful prediction models have also been established for different bioprocess products like the antibiotic polymyxin B (Eliasson Lantz et al, 2006), 5-aminolevolunic acid (Rhee and Kang, 2007) and recombinant proteins (Clementschitsch et al, 2005;Mortensen and Bro, 2006). Furthermore, in the literature there are numerous examples of successful indirect prediction models of variables that are inherently non-measurable with the MWF technology, e.g., correlation of off-gas CO 2 concentration or glucose consumption with culture fluorescence (Skibsted et al, 2001;Hagedorn et al, 2003Hagedorn et al, , 2004Haack et al, 2004;Lee et al, 2005;Surribas et al, 2006b;Rhee and Kang, 2007). These models are based on correlations between the available analyte trajectories in the data set, e.g., the consumption of a substrate can be modelled based on signals related to cell growth.…”

Evaluation of wavelength selection methods for 2D fluorescence spectra applied to bioprocesses characterization

“…In this case, their concentrations can still be indirectly predicted using the signals of available fluorophores (Rossi et al, 2012). In the literature, there are several examples of successful indirect prediction models based on 2D fluorescence spectroscopy (Skibsted et al, 2001;Solle et al, 2003;Lee et al, 2005;Surribas et al, 2006;Rossi et al, 2012). However, the metabolism of a microorganism can shift during the cultivation as a result of environmental changes, such as sequential uptake of different substrates, availability of oxygen or accumulation of inhibitors.…”